US 20050267718 A1 Abstract A method for enhancing the allocation of fluid flow rates among a plurality of well bores in fluid communication with at least one subterranean reservoir is disclosed. An objective function and system equations are generated which utilize constraint violation penalties associated with soft constraints. The soft constraints are constraints which may be violated if necessary to arrive at a feasible solution to optimizing the objective function and the system equations. The fluid flow rates are then allocated among the well bores as determined by the optimizing of the objective function and system equations. Fluid flow rates among well bores, particularly those exhibiting similar fluid characteristics, may be related to one another. Initial flow rates of components (oil, gas, and water) and pressures in the well bores may be determined by an initial simulation run. Then additional component flow rates may be estimated by scaling the original component flow rates based upon changing pressure draw downs in the well bores.
Claims(19) 1. A method for enhancing the allocation of fluid flow rates among a plurality of well bores in fluid communication with at least one subterranean reservoir, the method comprising:
(a) simulating fluid flow of a fluid containing multiple components in at least one subterranean reservoir and in a plurality of well bores which are in fluid communication with the at least one subterranean reservoir; (b) selecting production constraints including at least one soft constraint wherein the at least one soft constraint may be violated; (c) generating system equations including component flow rate equations corresponding to the simulated fluid flow in the well bores and constraint equations including at least one soft constraint equation associated with the at least one soft constraint, the at least one soft constraint equation including a constraint violation penalty (CVP) which allows the at least one soft constraint equation to satisfy the soft constraint; (d) generating an objective function corresponding to the fluid flow in the well bores and to the constraint violation penalty; (e) optimizing the objective function utilizing an optimizer and the system equations to determine an enhanced allocation of fluid flow rates among the plurality of well bores wherein the at least one soft constraint may be violated if necessary to achieve a feasible solution to the optimization; and (f) allocating the fluid flow rates among the plurality of well bores as determined in step (e). 2. The method of
the production constraints include a plurality of soft constraints which may be violated; the system equations include a plurality of soft constraint equations corresponding to the soft constraints, each of the soft constraint equations including a respective constraint violation penalty (CVP) which allows that soft constraint equation to satisfy the respective soft constraint; and the objective function corresponds to the fluid flow in the well bores and to the constraint violation penalties; wherein the soft constraints may be violated if necessary to achieve a feasible solution to the optimization. 3. The method of
the soft constraints are prioritized as to the difficulty to which the soft constraints are to be violated. 4. The method of
weighting scale factors are associated in the objective function with constraint violation penalties of respective soft constraint equations and are weighted in accordance with the prioritization of the soft constraints associated with the respective soft constraint equations to make the higher priority soft constraints more difficult to violate. 5. The method of
the objective function comports with the mathematical expression: where OBJ=objective to be optimized; i=number of fluid components in the fluid; w_{1}=weighting scale factor for production of the i^{th }fluid in a well bore; j=the number of well bores; q_{ij}=quantity of the i^{th }component produced by the j^{th }well; k=number of constraint violation penalties associated with the soft constraints; w_{k}=weighting scale factor for the k^{th }constraint violation penalty; and CVP_{k}=k^{th }constraint violation penalty. 6. The method of
the objective function comports with the mathematical expression: where OBJ=objective to be optimized; i=number of fluid components in the fluid; w_{i}=weighting scale factor for production of the i^{th }fluid component in a well bore; j=the number of well bores; q_{ij}=quantity of the i^{th }component produced by the j^{th }well; k=number of constraint violation penalties associated with the production constraints; w_{k}=weighting scale factor for the k^{th }constraint violation penalty; and CVP_{k}=k^{th }constraint violation penalty. 7. The method of
the soft constraints are prioritized as to the difficulty to which the soft constraints are to be violated; and the weighting scale factors w_{k }associated with the constraint violation penalties CVP_{k }of the respective soft constraint equations are weighted in accordance with the prioritization of the soft constraints to make the higher priority soft constraints more difficult to violate. 8. The method of
comparing characteristics of fluid flow in at least two well bores and if the characteristics are within a predetermined range of one another, then relating the fluid flow rates of the at least two well bores together by generating rate relating equations in the system equations so that the at least two well bores will have related allocated flow rates. 9. The method of
the well bores include a plurality of completion elements and the at least one subterranean reservoir includes a plurality of reservoir elements which are in fluid communication with the completion elements; the step of simulating fluid flow includes determining pressures in the reservoir elements and in the completion elements and includes determining the corresponding component fluid flow rates in the completion elements due to the pressure draw down between the reservoir elements and the completion elements; and the component flow rate equations are generated from component rate data points which are created by scaling and summing the component fluid flows in the completion elements of each well bore based upon the component fluid flow rates determined in the simulation of fluid flow and in relation to the changing pressure draw down between the reservoir and completion elements. 10. The method of
the component rate data points are generated utilizing the following mathematical expression: where q_{pT} ^{*}=new total quantity of flow from a well bore; ^{n}comp=number of completion elements in a particular well bore; ΔP_{i}=original pressure draw down of the i^{th }completion element; c=change in pressure draw down from the original simulated pressure draw dawn for a completion element; and q_{pi}=original simulated quantity of component flow from the i^{th }completion element. 11. A method for enhancing the allocation of fluid flow rates among a plurality of well bores in fluid communication with at least one subterranean reservoir, the method comprising:
(a) simulating fluid flow of a fluid containing multiple components in a plurality of well bores and in at least one subterranean reservoir, the well bores including a plurality of completion elements and the at least one subterranean reservoir including a plurality of reservoir elements which are in fluid communication with the completion elements, and determining pressures in the reservoir elements and in the completion elements and determining the corresponding component flow rates in the completion elements due to the pressure draw down between the reservoir elements and the completion elements; (b) generating component rate data points for the well bores over a range of fluid flows by scaling and summing the component fluid flows in the completion elements based upon component flow rates determined in step (a) and changing pressure draw downs between the reservoir and completion elements; (c) generating component flow rate equations for the well bores based upon the data points for the respective well bores; (d) generating constraint equations corresponding to production constraints; (e) generating an objective function corresponding to the fluid flow in the well bores; (f) optimizing the objective function utilizing an optimizer and the constraint and component flow rate equations to determine an enhanced allocation of fluid flow rates among the plurality of well bores; and (g) allocating the fluid flow rates among the plurality of well bores as determined in step (f). 12. The method of
generating piecewise linear functions from the data points for each of the well bores; and generating the component flow rate equations from the piecewise linear functions. 13. The method of
the component flow rate equations include binary variables to describe the piecewise linear function; and the optimizing step includes using mixed integer programming. 14. The method of
the flow rates between at least two of the well bores is related to one another. 15. The method of
the optimizer utilizes at least one of linear programming and mixed integer programming in optimizing the objective function. 16. The method of
the optimizer utilizes mixed integer programming in optimizing the objective function. 17. The method of
the component rate data points are constructed utilizing the following mathematical expression: where q_{pT} ^{*}=new total quantity of flow from a well bore; ^{n}comp=number of completion elements in a particular well bore; and ΔP_{i}=original pressure draw down of the i^{th }completion element; c=change in pressure draw down from the original simulated pressure draw dawn for a completion element; and q_{pi}=original simulated quantity of flow of a component from the i^{th }completion element. 18. A method for enhancing the allocation of fluid flow rates among a plurality of well bores in fluid communication with at least one subterranean reservoir, the method comprising:
(a) simulating fluid flow of a fluid containing multiple components in at least one subterranean reservoir and in a plurality of well bores which are in fluid communication with the at least one subterranean reservoir; (b) generating rate relating equations between at least two of the well bores; (c) generating an objective function corresponding to the fluid flow in the well bore; (d) optimizing the objective function utilizing the rate relating equations to determine an enhanced allocation of fluid flow rates among the plurality of well bores with the at least two well bores having related flow rates; and (e) allocating the fluid flow rates among the plurality of well bores as determined in step (d). 19. The method of
fluid flow characteristics are compared between the at least two well bores; and the generation of the rate relating equations is triggered if the compared fluid flow characteristics are within a predetermined range of one another. Description The present invention relates generally to methods for controlling hydrocarbon production from a field of wells, and more particularly, to methods for optimizing production by enhancing fluid flow rate allocations among the wells. Field scale optimization is known which attempts to optimize or enhance the production of production fluids, including hydrocarbons, from a field containing one or more subterranean reservoirs. Wells or well bores connect the reservoirs with surface facilities which collect and process the captured production fluids. Typically, these production fluids include the components of oil, gas and water. Chokes or flow control devices are used to adjust the allocation of flow rates among the well bores in a field. The relative quantities and ratios of production of the different components of oil, gas and water for an individual well bore can be controlled by adjusting a choke to change the pressure in a well bore. Surface facilities are needed to produce and process the production fluids. These facilities may include apparatus such as separators, pumps, storage tanks, compressors, etc. Ideally, the capital expenditures on these facilities are minimized by employing the smallest and least expensive surface facilities possible. However, fluid handling capacity should be sufficiently large so as not to unduly limit the production rate of the economically desirable oil and/or gas. Hence, the allocation of fluid flow in the well bores is ideally optimized to maximize monetary return while meeting production constraints such as those imposed by the fluid handling capacities of the surface facilities. Optimization techniques are used predict the optimal allocation of fluid flows in well bores for a given set of production constraints. First, a reservoir simulator is used to mathematically model the flow of fluids throughout a field including the reservoirs and well bores. The simulated flow is used to establish component flow rate curves or rate equations for each well bore which describe how the flow rate of one component, such as water, relates to the flow rate of another component, i.e., oil. Typically, an objective function is created which seeks to optimize an objective such as maximizing oil production or minimizing water production. The objective function incorporates the flow rates from the well bores which are predicted by the reservoir simulation. A set of production constraints, such as oil production targets or gas or water production limitations for the field, are specified. Constraint equations are generated to meet these production constraints. The fluid flow among the well bores must adhere to these production constraints. The objective function is then optimized by a subroutine, referred to as an optimizer, to determine the optimal allocation of flow rates among the well bores. The optimizer utilizes the well bore component flow rate equations and constraint equations in the optimization process. A first shortcoming of typical field scale optimization schemes is that feasible solutions to an optimization may not be possible for specified production constraints. For example, a certain level of oil production may be desired while not producing more than a specified quantity of water. A feasible solution to the objective function with this set of constraints may not be possible. In this event, one or more of the constraints must be adjusted and the reservoir simulator and optimizer run again to determine when a feasible solution is possible. Such iterative runs in solving numerous optimizations of the objective function are computationally intensive and undesirable. A second problem in some optimization schemes is that while a feasible solution to the optimization of the objective function may be achieved, the results may not be practical. For example, in a first run or time step, the optimizer may determine that a first well bore should produce at a high level while a second well bore is substantially closed down. In the next time step, the optimizer may suggest that the second well bore produce at a high level while the first well bore is substantially shut down. Therefore, production from the well bores may oscillate if the suggested allocations from the optimizer are followed. Generally, it is more practical if the production from well bores having similar fluid flow characteristics are at a consistent level. This would minimize the oscillations in production from the related well bores over time steps. A third shortcoming is that creating component flow rate curves or equations for the production of fluids from a well bore can be computationally intensive. One method of calculating these rate curves is to create a sub model of the well bores and surrounding reservoirs and iteratively solve for the production rates of the components, i.e., oil, gas and water, as the chokes are opened and the pressure draw downs between the reservoirs and the well bores are increased. Typically, several Newton iterations must be performed to produce each data point relating the production of one component relative to another component for a given pressure draw down in a well bore. Again, the pressure draw down in a well bore is related to how open is a choke controlling the well bore. This process is repeated many times until enough data points, perhaps as many as 30-50 data points, have been calculated such that an overall flow rate curve or equation can be developed. The optimizer then uses the rates curves or equations during the optimization of the objective function. Generating data points using these many Newton iterations to create rate curves or equations is computationally costly. The present invention provides solutions to the above described shortcomings of conventional field scale optimization schemes. First, an objective function and associated constraint equations are generated which can be solved in a single run of an optimizer to produce a feasible solution. Second, constraint equations may be created which requires the rates of production from similar well bores to be related to prevent significant oscillation of well rates between time steps of a reservoir simulation. Finally, an efficient method of generating well bore component flow rate curves or equations relating production rates between fluid components of a well bore is described. The present invention includes a method for enhancing the allocation of fluid flow rates among a plurality of well bores in fluid communication with at least one subterranean reservoir. Fluid flow is simulated, using a numerical reservoir simulator, in at least one subterranean reservoir and in a number of well bores in fluid communication with the subterranean reservoir. Component flow rate equations are generated from the simulated flow in the well bores. Production constraints are selected with at least one of the production constraints ideally being a soft constraint which may be violated if necessary during an optimization process to provide a feasible solution. Constraint equations corresponding to the production constraints are also generated. An objective function is generated which corresponds to the fluid flow in the well bores. The objective function may also include constraint violation penalties which correspond to the soft constraints and soft constraint equations. The objective function is then optimized utilizing the component flow rate equations and the constraint equations to determine an enhanced allocation of fluid flow rates among the well bores. If necessary, soft constraints may be violated to achieve a feasible solution to the optimizing of the objective function. The presence of the constraint violation penalties allows the soft constraints to be violated while still satisfying a corresponding constraint equation. The fluid flow rates are then allocated among the well bores as determined by the optimizing of the objective function. The soft constraints may be prioritized as to which of the soft constraints should be most difficult to violate if necessary to achieve a feasible solution to the optimization of the objective function. Weighting scale factors may be associated with the constraint violation penalties in the objective function. The weighting scale factors may be weighted in accordance with the prioritization of the soft constraints to make higher priority soft constraints more difficult to violate than lower priority soft constraints. Flow rates between select well bores may have their flow rates related. In particular, well bores exhibiting similar flow characteristics, such as gas-to-oil ratio (GOR) or water-to-oil ratio (WOR), may have their well rates related to one another. Again, constraint equations can be generated for these related well bore flow rates. The enhanced allocation of flow rates among the related well bores will then be related or tied to one another. In another aspect of this invention, the simulated well bores include a plurality of completion elements and the reservoir or reservoirs include a plurality of reservoir elements. The reservoir simulator is run to determine pressures in the reservoir elements and in the completion elements and to determine fluid flows in the completion elements of at least two components, i.e., oil and water, due to the pressure draw down between the reservoir elements and the completion elements. Fluid flow component rate data points are then generated over a range of fluid flows for each well bore. The data points are ideally generated by scaling and summing the fluid flows in the completion elements based upon the component flow rates determined by an initial simulator run and in relation to an incremented range of pressure draw downs between the reservoir and completion elements. It is an object of the present invention to provide a method wherein an objective function is created which includes at least one constraint violation penalty corresponding to a soft constraint which allows the objective function to be optimized wherein the soft constraint may be violated if necessary to arrive at a feasible solution for the optimization. It is another object to generate an objective function which incorporates weighted constraint violation penalties which may be appropriately weighted so that soft constraints may be violated in a prioritized order. It is yet another object to relate production rates of well bores in an optimization so that the flow rates among those well bores will have related flow rates after an optimization has been performed resulting in limited flow rate oscillations of those well bores between time steps in a reservoir simulation. It is still another object to generate component flow rate equations which are generated by scaling component flow rates in individual completions elements based upon flow rates originally determined in a reservoir simulation run and a range of changing pressure profiles within the well bores. These and other objects, features and advantages of the present invention will become better understood with regard to the following description, pending claims and accompanying drawings where: FIGS. 5A-D illustrate oil, gas and water production from individual completion elements of a well bore as the pressure profile in a well bore is allowed to increase due to the simulated closing of a well bore choke; A. Overview Chokes or well control devices 54, 56, and 60 are used to control the flow of fluid into and out respective well bores 30, 32 and 34. As will be described more fully below, chokes 54, 56 and 60 also control the pressure profiles in respective well bores 30, 32 and 34. Although not shown, well bores 30, 32 and 34 will fluidly connect with surface facilities such as oil/gas/water separators, compressors, storage tanks, pumps, pipelines, etc. The rate of flow of fluids through well bores 30, 32 and 34 may be limited by the fluid handling capacities of these surface facilities. A reservoir simulator is used to model the fluid flow in field 50 which includes the reservoirs and well bores (step 110). Generally, such a reservoir model will include thousands or even millions of discrete elements to carry out a numerical simulation. These discrete elements comprise reservoir elements and well bore elements. The well bore elements include specific completion elements which transfer fluid back and forth between adjacent reservoir elements and other well bore elements which are in fluid communication with the choke and the surface facilities (not shown). Initial and boundary conditions are specified on the field model. These initial and boundary conditions include, by way of example and not limitation, the initial pressures and flow rates in the reservoir elements and well bore elements, fluid compositions, viscosities, etc. Next, a simulation run (step 120) is performed on the field model to calculate reservoir and fluid flow characteristics for a time step. In particular, fluid flow rates between the reservoirs and the well bores are determined as are the pressures in the reservoir and well bore elements. Producing well bores will receive producing fluids from the reservoirs, including oil, water and gas, which are delivered to the surface facilities of the field. Injection wells may be used to pressurize one or more of the reservoirs and/or to dispose of water. Also, gas may be injected into the well bores to provide gas assisted fluid production. Those skilled in the art will appreciate that many other operations affecting production may be modeled with a reservoir simulator and these operations are included within the scope of this invention. Component fluid flow rates may be determined in terms of oil, gas and water flow. Alternatively, the fluid components for which flow is to be optimized could be compositional components such as light (C_{3}-C_{4}), medium (C_{5}-C_{8}) and heavy (>C_{9}) hydrocarbons. By way of example, and not limitation, other possible component combinations might include non-hydrocarbon components such as H_{2}S and CO_{2}. Component flow rate equations for each of the well bores are next calculated (step 130.) These component flow rate equations describe the estimated flow of one fluid component relative to that of another fluid component over the anticipated range of flow rates for a well bore. Physically, the chokes on the well bores may be opened or closed to increase or decrease the overall fluid output or input relative to a well bore. Because of changing pressure profiles in the well bores, the relative ratios of oil, gas and water produced from a well bore may change with the opening and closing of a choke. Examples of component flow rate curves for a well are shown in In the preferred embodiment of this invention, a “quick-rates” method is used to generate individual component rate data points which can then be used to quickly construct graphs or generate component flow rate equations. More details on the “quick-rates” method will be described below. Those skilled in the art will appreciate that other methods may be used in generating estimates of how the production of one component versus the rate of production of another component may vary over the overall output range of a well bore. A user will specify production constraints (step 140) to be used in conjunction with the field model. By way of example and not limitation, examples of production constraints include (1) producing oil at a target level; (2) producing gas at a target level; (3) limiting gas production below a predetermined limit; (4) limiting water production below a predetermined limit; (5) limiting water injection to an amount related to the water produced from the well bores; and (6) limiting gas injection above a predetermined limit to provide gas assisted lift. Further, these targets and limitations may be combined or scaled relative to one another as well. The production constraints may include hard or soft constraints. Hard constraints are constraints which will not be allowed to be violated. Soft constraints are constraints which may be violated if necessary to produce a feasible solution to an optimization problem. Optionally, the order in which the soft constraints are preferably allowed to be violated, if necessary to achieve a feasible solution, may also be specified. Another aspect of the present invention includes optionally specifying (step 150) whether the well bore flow rates of certain well bores are to be related. For example, well bores having similar fluid characteristics such as gas-to-oil ratio (GOR) or water-to-oil ratio (WOR), may be related to one another. The relating of production rates between well bores will insure that rates of production (or injection) between these well bores will not arbitrarily oscillate between time steps. Constraint equations are then generated (step 160) from the production constraints and the related well bore rates. Hard constraint equations are created for those constraints which are not allowed to be violated. Soft constraint equations corresponding to the soft constraints are generated which include constraint violation penalties. The constraint violation penalties allow the soft constraint equations to be satisfied even when the soft constraints must be violated so that an optimization may produce a feasible solution. The generation of this set of constraint equations will be described in further detail below. An objective function is created in step 170 which seeks to optimize an objective, such as oil production from field 50. The objective function ideally includes the component flow rates of the well bores and also the constraint violation penalties associated with the soft constraint equations. Weighting scale factors may be associated with the soft constraint penalties in the objective function. By appropriately weighting these weighting scale factors, the order in which related soft constraints may be violated, may be prioritized. The objective function is then optimized (step 180) by an optimizing subroutine (optimizer) to produce an optimized allocation of fluid flow rates among the well bores. The optimizer uses the component flow rate equations calculated in step 130 and the constraint equations set up in step 160 to optimize the objective function. The optimized fluid flow rates, and other fluid flow characteristics determined from the optimizer such as constraint violation penalties, may then be allocated among the well bores and reservoir (step 190). These optimized flow rates and characteristics may then be imposed (step 200) as initial/boundary conditions in the next iterative time step in the reservoir simulation. Steps 120-200 are then repeated to provide enhanced field scale production over many time steps until a satisfactory period of time has elapsed and the simulation is then ended. More details on the above aforementioned steps will now be described. B. Creation of the Objective Function and Constraint Equations
A linear programming (LP) system is a set of linear equations and linear constraints. A mixed integer programming (MIP) system is a set of linear or non-linear equations and constraints. In the present invention, preferably a MIP system augments a LP system when a set of non-linear equations or constraints, represented by piecewise linear functions, needs to be solved to achieve an optimized objective. An open source software package, which uses LP and MIP techniques, is used in this exemplary embodiment to optimize the objective function. In particular, the present invention uses a package entitled LP-Solve, which is available from http://packages.debian.org/stable/math/lp-solve. An alternative commercial solver is also utilized entitled XA which is available from Sunset Software Technology Corporation, of San Marino, Calif. Those skilled in the art will appreciate that other commercial LP/MIP optimizer packages may be used to optimize the objective function using fluid flow rates and constraint conditions. The constraint equations, component flow rate equations, and the objective function are input into the optimizer. The optimizer then outputs a feasible solution to the optimization problem including enhanced allocation of well bore flow rates. Values for the violation of any soft constraints necessary to achieve a feasible solution to the optimization are also ideally output. A user may then make appropriate changes to production constraints or to the capacity of surface facilities to reflect the value of the violation of the soft constraints. An extrema of an objective function is sought. A simple LP system may have the following form:
In one embodiment of this invention, the main variables are well bore rates. That is, the rates at which components of fluid production, i.e., oil, water and gas, are produced from a well bore. Component flow rate equations are preferably generated using a “quick rates” method which will be described below. The component rate equations describe how much of one component is transported through a well bore as compared to another fluid component. The rates of production of the components may remain linear with respect to one another or may be non-linear over the potential range of well bore production outputs. The present invention ideally handles nonlinear scaling between component or phase rates through piecewise linear functions by formulating the system as a MIP problem. Production constraints are set up as hard constraints, which are not allowed to be violated, and/or as soft constraints, which are allowed to be violated when necessary to achieve a solution. The constraints may include target objectives and production limitations. The objective function is setup from information provided by a user.
In general, the objective function comports with the mathematical expression:
A more specific exemplary objective function for the LP/MIP system might consist of the weighted sum of total production rates of oil, water and gas for a selected set of well bores. In the present invention, the objective function may also include constraint violation penalty variables (CVP_{k}) to accommodate the use of soft constraints. A typical objective function may be expressed in the following mathematical form:
The weighting scale factors wi or well rate parameters may be specified by a user. For example, a user might specify:
These weighting scale factors correspond to the maximization of oil production rate while trying to minimize gas and water rates. In this case, the objective function is incremented by 1.0 for each stock tank barrel/day (STB/D) of oil produced (w_{oil}=1.0) and penalized by 0.2 for every million standard cubic feet/day (MSCF/D) of gas and 0.1 for every STB/D of water produced. In this case, the units of the objective function are a combination of STB/D and MSCF/D units. Normalization of the objective function components is ideally carried out to render the objective function non-dimensional. Another preferred way of handling this unit mismatch in the objective function is to make use of economical information, if available. For example, if oil revenues are 22$/STB/D, gas revenues are 3$/MSCF/D and every STB/D of water costs $3.5 to handle, then:
In this case, the units of the objective function are monetary ($) and are consistent. It is preferred to scale the weighting scale factors so that w_{o }is 1.0, hence the previous well rate parameter values would be normalized by 22.0 to give:
Constraints may be based on physical limitations such as well production limits, injection rate limits or gas lift rate limits. Alternatively, constraints may be determined to meet engineering preferences such as production/injection targets for a group of wells. Other constraints by way of example and not limitation might include Gas to Oil Ratios (GOR), Water to Oil Ratios (WOR), and constraints on a subset of wells or completions. The LP/MIP system constraints are classified as hard and soft constraints. For example, hard constraints may be imposed on a pair of wells such that the combined maximum oil production is 5,000 STB/D. These hard constraints are translated into the following LP/MIP constraints:
Soft constraints are constraints that are allowed to be violated if-and-only-if there is no other way to honor the soft constraints while obtaining a feasible solution for the system. Ideally, this violation of constraints will be the minimum possible necessary for obtaining a solution. Constraint violations may occur when the system has conflicting limits/targets. Consider the following situation where the field has constraints including an oil production target and a water handling limit on a group of wells as follows:
There might, and most probably, will be a point in the simulation where the group of wells will not be able to produce 7,500 STB/D of oil without producing more than 5,000 STB/D of water. Wells tend to produce more water as they age or mature. In such a case, the optimizer will not report a no-solution but instead will allow the violation of one of the soft constraints. Preferably, a flag will be raised indicating that the constraint has been violated. Which constraint is chosen to be violated first may be determined by the user as well in this preferred embodiment of this invention. These oil target and water limit conditions are translated into the following three soft constraint equations:
Constraint violation penalty CVP_{k }variables are appended to the objective function:
subject to: w_{k}>0 where w_{k }is the k^{th }weighting scale factor associated with the k^{th }constraint violation penalty; and CVP_{k}>0 where CVP_{k }is the k^{th }constraint violation penalty which is associated with the k^{th }constraint equation. Note that this setup forces the CVP variables to be zero since they have negative weights in the objective function which makes them equivalent to hard constraints whenever they can be met, i.e., when oil production is equal to 7,500 STB/D and water production is less then 5,000 STB/D. Suppose the reservoir conditions are such that in order to produce 7,500 STB/D of oil, 5,100 STB/D of water has to be produced. In this case, there are two options:
Whether the LP/MIP system chooses to scale back production or meet the water limit depends on the coefficients or weighting scale factors w_{k }of the CVP_{k }variables. Suppose the water capacity limit is absolute and that the oil production is allowed to be scaled back to meet the water limit. In this case, suppose w_{1}=1, w_{2}=1 and w_{3}=2 which corresponds to constraint-3 (water production limit) having more priority than the other two constraints (oil production target). Note that the weighting scale factor w_{3 }is given greater weight than the other two weighting scale factors w_{1 }and w_{2 }associated with the oil production. When the well rates are scaled back to meet the water production limit, suppose the oil production drops to 7,400 STB/D when water production is exactly 5,000 STB/D. In this case, CVP_{1 }will have to be non-zero to satisfy constraint 1, to be exactly CVP_{1}=100. In this setting, the LP/MIP system will choose to scale back the rates rather than produce more water due to the specific values of CVP coefficients w_{k}. The objective function entries will appear as follows for these two cases. If the oil production target is disregarded and oil production is allowed to be scaled back to meet the water limit, then:
If the oil production target is enforced but the limit on the water production limit is allowed to be violated, then:
Since, everything else being the same, scaling back rates results in a higher objective function value (+100), the LP/MIP optimizer will prefer to scale back the rates. The same approach may be used to handle n soft constraints and put them in a desired priority order of violation. If the order in which the soft constraints are to be violated is not specified and remains unprioritized, then all of the weighting scale factors w_{k }are equal and no preference is given as to which constraint is allowed to be violated first. In this event, w_{1}=w_{2}=w_{3}=1. Alternatively, a first soft constraint may be given the lowest priority, a second soft constraint is given a slightly higher priority, and a third soft constraint is given the highest priority. In this exemplary embodiment of the invention, the weighting scale factors w_{i }are then given values corresponding to 10×10^{P }where p is order of priority in which the soft constraints may be violated. For example,
The general equation for the objection function is:
The objective function with weighting scale factors then becomes:
Preferably, these coefficients are normalized to give values of between 0 and 1. The normalization is partially based upon the potential range of a constraint violation penalty.
Or, since CVP_{min }is always zero:
CVP_{k }parameters are optimized along with the other parameters in the optimization system (production/injection rates). Since any positive value of CVP_{k }imposes a penalty through the objective function, the system tries to keep CVP_{k }values as zero. CVP_{k }gains a positive value if and only if there is no other way to achieve a feasible solution. Note that if there are no conflicting objectives for optimization, all of the CVP variables will be zero and soft constraints will be equivalent to hard constraints. The operators used with the soft constraints are translated into LP/MIP equations as follows:
LP/MIP systems are strictly mathematical and thus have no notion of the physics underlying the variables, equations and constraints. Therefore, in some cases, the LP/MIP results, although mathematically sound, may make little practical sense. Such a case may occur when the LP/MIP optimizer decides to significantly choke back only one well bore in a group of well bores that all have insignificant differences in their properties. This might result in large rate oscillations for individual wells between time steps. To prevent such an occurrence, the present invention provides the option that well rates of well bores with close characteristics be related. If it is determined that the well rates should be related, in addition to the existing constraint equations, further constraints equations that relate certain well bore flow rates are setup. For example, if well bores which have fluid characteristics which are within a predetermined range of one another, such as gas-to-oil ratios (GOR) and/or water-to-oil ratios (WOR), then the flow rates of these well bores may be related. Similar to the soft constraint equations described above, these rate relating equation may have weighting scale factors which are close to one another and include constraint violation penalties. Referring now to
This means that when q_{1}=q_{1} ^{*}, q_{2 }needs to be in the range |q_{2} ^{*} _{min}, q_{2} ^{*} _{max}|. The function f is a simple linear function as shown in The present invention allows a user to change a threshold value t, however, t=1.0 should work for most cases. With this setting, given t=1.0, a well with GOR_{2}=0.0 will not be related, and will have an independent rate scaling factor, whereas on the other extreme when GOR_{2}=GOR_{1}, the shaded area in Another way to relate flow rates is through scaling flow rates in a group of well bores by the same factor. For example, the injection rates of all the injectors in a first injector well group and the production rates of all the producers in a first production group of well bores may be related. This relation is not based on GOR or WOR in this case; the relation simply implies that when the rate of a well bore is scaled by a factor, the other wells in the related group will be scaled with the same factor. For instance, if a well bore in a first production group needs to cut production by half (to satisfy another constraint perhaps), then all the well bores in a first production group will cut production by half. Ideally, the default for this relation is to have less weight in the LP/MIP system than the specified constraints. This means that rate-relations may be broken for the sake of satisfying the constraints. Parameters can be used to determine the relative weights of the constraints and rate-relations in the LP/MIP system. The smaller (more negative) these coefficients, the more influence these coefficients will have on the system. C. Generation of Rate Curves and Component Flow Rate Equations
The following “quick rates” method is preferably used in generating fluid flow component flow rate curves and equations. A rate curve relates how the production of one component compares with the production of another. For example, as a choke or valve is opened on a well, oil, water and gas production will generally increase. The increase between any two of the components may be linear or non-linear over the range of overall fluid production. Referring again to The quick-rates method utilizes the fact that at a fixed point in time, production from individual completion elements is generally linearly proportional to pressure draw down. Pressure draw down is the pressure differential between the pressure in a well bore completion element and adjacent reservoir elements. It is this pressure differential which drives fluids into and out of the completion elements during respective production and injection operations. Using a number of different pressure draw down profiles for each well bore, a set of data points is generated. Then, a piecewise linear function that best fits these points is ideally constructed. A component flow rate equation is then generated from this piecewise linear function which is to be used by the optimizer. The oil-water total component flow rate curve is piecewise linear, which is not a linear function. FIGS. 5A-D show the flow rates of individual completion elements for four different overall production outputs for a well bore. Also shown are the pressure profiles for the reservoir and well bore elements for these different production rates. FIGS. 5A-D illustrate cases where oil production is being sequentially reduced, such as occurs when a well head choke valve of a well bore is being closed. Note as oil production is reduced, water production is reduced until almost no water is produced. While the rate of production is decreased, the well bore pressure profile of the well bore will increase. The pressure profile of the reservoir is assumed to remain constant at a given time step. This will result in the pressure draw down in the well decreasing as the well bore pressure profile increases toward the reservoir pressure profile. Note that the pressure at deeper completions will be greater than at shallower depth completions due to pressure head/gravity effects. Consequently, pressure draw down will be lower at greater depths where denser water underlies less dense layers of oil and gas. The present invention exploits the linear rate scaling for individual well bore completions. The total production rate of component p, i.e., oil, water or gas, from a well is the sum of rates from its flowing completions:
The baseline flow rate of each component at each individual completion is extracted from the reservoir simulation run at a particular time step and well production level. It is assumed that at a fixed point in time the completion rate for each individual well completion element is linearly proportional to the pressure draw down. Thus, if the pressure draw down in a well is reduced by an amount, c, individual completion rates will be scaled back accordingly and the new total rate will be given by:
Thus, the amount of pressure shift, c, required to reduce the well oil rate from q to q^{* }is given by:
This pressure shift dictates a parallel shift in the well bore pressure profile, as illustrated in
Piecewise linear functions are generated which best represent these data point sets generated by the “quick-rates” method for each of the well bores. The piecewise linear functions include a number of line segments and breakpoints. The number and location of the breakpoints are ideally selected using a least squares fit to the data set generated by the “quick-rates” method. In this exemplary embodiment, a Levenberg-Marquardt least squares fit method is preferably used to locate breakpoints. Those skilled in the art will appreciate other curve or equation generating techniques may be used to represent the generated data points which is to be used by the optimizer. Referring now to Least square methods, such as the Levenberg-Marquardt, require the derivatives of this function, y, be determined with respect to the parameters, α. These derivatives are:
In the preferred embodiment, ideally an appropriate number of breakpoints as well as their optimum locations are determined. The algorithm shown in The first step is to start with a linear function (i.e. a single segment, two end points, hence i=2. The x_{i} ^{2 }for this linear function is calculated (x_{2} ^{2}). Then a break point is added to the linear function making it a piecewise linear function with two segments and three end points (i=i+1, i.e., i=3). The breakpoint coordinates is optimized for minimum x_{i} ^{2}. If the fit is improved by more than a factor of f from the initial fit, then a new breakpoint is added and the process is repeated until the improvement is not significant. This algorithm keeps adding more breakpoints only if this improves the fit by the fraction f. A better fit can be made by decreasing the value of f at the expense of having a larger number of segments. This approach is generally robust. A check may be made to make sure that the breakpoints are always in the feasible region (first quadrant). This is ensured by penalizing (P) the solutions that fall into infeasible areas, as shown in
Incorporation of piecewise linear curve to the LP setup requires the introduction of binaries, additional continuous parameters and some constraints. Following is the set of equations and variables that need to be added: Break points:
Replace rate term with:
Add constraints:
Here, q is the dependent rate and q_{1 }is the controlling rate. Now it will be demonstrated why such a setup results in the correct behavior with a simple piecewise linear function with the two segments. Suppose the function appears as in The binary y indicates the segment that x belongs to. In this case, y_{1 }should be one and y_{2 }should be zero. First, check to see if y_{2 }can ever be one. If y_{2 }was one, then y_{1 }has to be zero, which means z_{1 }is zero and z_{2 and z} _{3 }are non-zero. However, if z_{2 }and z_{3 }are non-zero, x=15 for the second equation can never be satisfied, thus y_{2 }cannot be 1. Thus if y_{1 }is one, then solving for z obtains:
Incorporation of the equations and variables in equation 24 force the LP/MIP optimizer to honor the component flow rate curves. While in the foregoing specification this invention has been described in relation to certain preferred embodiments thereof, and many details have been set forth for purpose of illustration, it will be apparent to those skilled in the art that the invention is susceptible to alteration and that certain other details described herein can vary considerably without departing from the basic principles of the invention. Referenced by
Classifications
Legal Events
Rotate |